The aim of this study was to compare the clinical usefulness of ultrasmall superparamagnetic iron oxide (USPIO) MR contrast media (Sinerem, Guerbet Laboratories, Aulnay-sous-Bois, France) with precontrast MRI in the diagnosis of metastatic lymph nodes in patients with head and neck squamous cell carcinoma, using histology as gold standard. Eighty-one previously untreated patients were enrolled in a multicenter phase-III clinical trial. All patients had a noncontrast MR, a Sinerem MR, and surgery within a period of 15 days. The MR exams were analyzed both on site and by two independent radiologists (centralized readers). Correlation between histology and imaging was done per lymph node groups, and per individual lymph nodes when the short axis was > or = 10 mm. For individual lymph nodes, Sinerem MR showed a high sensitivity (> or = 88%) and specificity (> or = 77%). For lymph node groups, the sensitivity was > or = 59% and specificity > or = 81%. False-positive results were partially due to inflammatory nodes; false-negative results from the presence of undetected micrometastases. Errors of interpretation were also related to motion and/or susceptibility artifacts and problems of zone assignment. Sinerem MR had a negative predictive value (NPV) > or = 90% and a positive predictive value (PPV) > or = 51%. The specificity and PPV of Sinerem MR were better than those of precontrast MR. Precontrast MR showed an unexpectedly high sensitivity and NPV which were not increased with Sinerem MR. The potential contribution of Sinerem MR still remains limited by technical problems regarding motion and susceptibility artifacts and spatial resolution. It is also noteworthy that logistical problems, which could reduce the practical value of Sinerem MR, will be minimized in the future since Sinerem MR alone performed as good as the combination of precontrast and Sinerem MR.